<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Sep 18, 2013 at 3:10 PM, Michael Schwartzkopff <span dir="ltr"><<a href="mailto:misch@clusterbau.com" target="_blank">misch@clusterbau.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div class="im"><p style="margin:0px;text-indent:0px">> I'm still a bit unclear on how the cluster monitors the VIP resources.</p>
<p style="margin:0px;text-indent:0px">> Do I have exactly one stanza of totem interface, and set the bindnetaddr to</p>
<p style="margin:0px;text-indent:0px">> the heartbeat net?</p>
<p style="margin:0px;text-indent:0px">> How does the cluster monitor for a VIP on a dead interface?</p>
<p style="margin:0px;text-indent:0px"> </p>
</div><p style="margin:0px;text-indent:0px">The monitoring operation of a IP address resource issues a "ip a l dev (...)" command and looks if the IP address is still bound to the interface. Any failure (i.e. interface down, IP address vanished) results in an error of the monitoring operation and a reaction of the cluster.</p>
</blockquote><div><br></div><div>That's what I expected, but not quite what I'm seeing. For a test I brought down the resident interface for a VIP. The monitor noticed a problem with the VIP, but did not move the VIP to the other node. Specifically, I create a cluster with ha-node2, ha-node3; each with an Internal and External interface. I created a VIP "InternalIP" and bound it to the Internal interfaces. I determined which Interface the VIP was bound to, and brought the interface down via "ifdown". My cluster now reports an error:</div>
<div><br></div><div># pcs status</div><div>Last updated: Wed Sep 18 07:35:34 2013</div><div>Last change: Wed Sep 18 06:58:19 2013 via cibadmin on ha-node2</div><div>Stack: classic openais (with plugin)</div><div>Current DC: ha-node3 - partition with quorum</div>
<div>Version: 1.1.8-1.tos2-394e906</div><div>2 Nodes configured, 2 expected votes</div><div>2 Resources configured.</div><div><br></div><div><br></div><div>Online: [ ha-node2 ha-node3 ]</div><div><br></div><div>Full list of resources:</div>
<div><br></div><div> InternalIP<span class="" style="white-space:pre"> </span>(ocf::heartbeat:IPaddr2):<span class="" style="white-space:pre"> </span>Started ha-node2</div><div> ExternalIP<span class="" style="white-space:pre"> </span>(ocf::heartbeat:IPaddr2):<span class="" style="white-space:pre"> </span>Started ha-node2</div>
<div><br></div><div>Failed actions:</div><div> InternalIP_monitor_30000 (node=ha-node2, call=19, rc=7, status=complete): not running</div><div><br></div><div><end of pcs status output></div><div><br></div><div>and ifconfig reveals the interface I brought down is now back up, and the IP has changed to the VIP IP. I expected the IpAddr monitor to detect the interface was down, and move the VIP to the other node. The cluster did not move the VIP. </div>
<div><br></div><div>Any idea what happened? Did I misconfigure?</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<p style="margin:0px;text-indent:0px"> </p>
<p style="margin:0px;text-indent:0px">Additionally use a ping resource. That resoruce sends a ping to an IP address outside of the cluster. If the node receives the answer it can be pretty sure that the attached network works.</p>
</blockquote><div><br></div><div>A ping resource is starting to sound more attractive.</div><div><br></div><div>thanks again,</div><div>Jeff</div><div> </div></div><br><br></div></div>